Call path
Jev AI on OpenRouter
Call the Decisions API with the recorded ids typesafe/jev-latest or typesafe/jev-1.13, not a chat-completions client. Re-check the live card; this page only repeats handbook ids, dates, and prices.
Last updated
How do you call the Jev model on OpenRouter?
Call the Decisions API for a model id this handbook has recorded: typesafe/jev-latest or typesafe/jev-1.13. Do not point a chat-completions client at it. The price and context figures below are the ones stored in this site’s fact card. Re-check the live OpenRouter card before you ship, and follow OpenRouter’s docs for their HTTP path.
Model ids recorded on this site
The September 2026 channel map records two OpenRouter cards and no others. This page does not invent a newer slug, a price, or an endpoint.
- typesafe/jev-latest — the moving alias, parallel to the direct API’s
jev-latest. - typesafe/jev-1.13 — the version card. The fact card stores the slug
jev-1.13and a listing date of 2026-09-18. The marketed version string is Jev 1.13. A direct API response can name a build such asjev-1.13.0.
If OpenRouter adds or renames a card after those notes, the live card wins. Do not copy an id from a community post unless it matches one of the two URLs above or the current OpenRouter page.
Decisions API, not chat completions
Both cards are the Decisions API. A chat-completions SDK pointed at OpenRouter chat will not speak System One. You still send a state and a map of questions, and you still get typed answers: noul, choice, or score, with probabilities. The API quickstart shows the direct TypeSafe shape. The primitives page is the field reference. OpenRouter’s own docs are the source for the HTTP path on their side; this handbook does not print a guessed OpenRouter URL.
POST https://api.typesafe.ai/v1/systemone with model jev-latest. That host is not the OpenRouter host. Price and context stored here
The fact card, checked against the Jev 1.13 card when the channel map was written, records $0.042 per million input tokens ($42 per billion) and free output. Context is 32,000 tokens. Input is text only (string / JSON object / array). Claimed P50 latency is about 0.23s. Re-check the live OpenRouter card and docs.typesafe.ai before you budget. The pricing calculator only multiplies the rates stored here. It does not call OpenRouter.
Which path to use
| Need | Prefer |
|---|---|
| TypeSafe console keys and the native System One body | Direct api.typesafe.ai. See the API quickstart. |
| OpenRouter next to other providers | The two model cards above, Decisions API only. |
| Vercel AI Gateway model ids and dates | The September 2026 channel map. That post is where the Gateway changelog links live. |
TypeSafe AI is the company. Jev AI is the model. This site is an unofficial handbook and is not affiliated with either OpenRouter or TypeSafe AI.
Names people search next to OpenRouter
The fact card’s OpenRouter traffic list is sel-jev-rubric-judge (9.18B tokens), Waura (2.64B tokens), mirasim (1.82B tokens), JeVinci (1.31B tokens), clara jev triage (1.27B tokens). sel-jev-rubric-judge is one of those app labels. This GitHub catalog does not attach a repository URL to that label. jarrodwatts/jev-trader is in the GitHub directory: Kuru/Monad trading experiment with optional Jev buy/sell decisions; defaults to a mock model and dry-runs without a private key, but configured execution can place real orders. Repository lists are on Jev AI GitHub projects. Community posts are on Jev AI cases.
What this page will not guess
- A model id other than
typesafe/jev-latestandtypesafe/jev-1.13. - An OpenRouter HTTP path, header, or request body that is not already the System One shape documented for the direct API.
- A price or context length other than the fact card figures above.
- That an OpenRouter app name is a GitHub repository.